Oh dear. The newspapers this morning are reporting that
Autism 'could be triggered by very low doses of anti-depressants or other chemicals found in water supply'Here's the study. Young fish were exposed to a combination of three drugs, two antidepressants and an epilepsy med, for 18 days.
First off, this study was tiny with an effective sample size of just 6. Three tanks of fish got exposed to the drugs, and three control tanks didn't. There were multiple fish per tank, five in fact, but those are not five independent observations, because they shared a tank. That's just tiny for a drug trial, or any scientific study really.
Next, the drug doses were much higher than in the water supply. Levels of fluoxetine (Prozac) were 700 times higher than observed in drinking water, for carbamazapine it was 400 times higher. And that's based on the authors' figures for drinking water which they admit are "the highest observed concentrations from various systems". The authors defend this by saying that in drinking water there will be other related compounds, on top of the drugs themselves, adding up to a higher dose. OK - but 400, 700 times higher? We've no idea if that's realistic. They don't justify this number.
The drugged minnows had significantly increased expression of a set of 324 genes dubbed "autism genes" ("autism_ideopathic" in the paper.) I'm not going to get into the question of whether these really are autism genes in humans, or whether fish brains are a good of model of humans. Those are hard issues. But what's easy to see is that while this set of genes were apparantly increased, so were many others. It was not specific to 'autism genes'.
The autism genes were upregulated by an average factor of +1.621... but this was only slightly more than the "Parkinson's Disease genes" at +1.56 and the "Multiple Sclerosis" ones at +1.375. Meanwhile, "Bipolar Disorder" genes were down by -1.172. So if antidepressants in the water are causing autism, they're probably also causing (or preventing!) a lot of other problems too.
The authors note that only three of the gene sets were statistically significantly altered, but that doesn't mean those sets were special, this is the fallacy of treating differences in significance levels as evidence of significant differences.
Of 10 more specific "autism gene" sets that they also examined (in the same fish), all were increased by various amounts (+1.050 to +1.537), some of which were significant - but one of those was a set of genes previously reported decreased in autistics, not increased (it was the "synapse" genes from this study).
What these changes in gene expression mean is anyone's guess. Given the small sample size they could be just noise. If not, all they really show is that levels of psychoactive medications that are quite low, but much higher than in drinking water, have affect the brains of fish. We don't know what that effect means, for the fish, let alone humans.
Early life antidepressant exposure might cause autism. I don't know. Stranger things have happened. We know that fetal anticonvulsant exposure can do it but that's when mothers are actually taking the pills. It's one giant leap from that to traces in drinking water. It's the difference between falling off your chair and falling off the Empire State Building.
Michael A. Thomas, and Rebecca D. Klaper (2012). Psychoactive Pharmaceuticals Induce Fish Gene Expression Profiles Associated with Human Idiopathic Autism PLoS ONE
27 comments:
Now there are people bringing up the MMR vaccine in the comments to the news article. Will the lies and science illiteracy ever end? :(
Not being able to accurately define nor diagnose autism in humans someone claims to find it in fish. And then wonder why their profession is loosing respect.
Autistic fish. Now there is something worth seeing. I wonder if they have delays starting to speak and the like?
I wonder if autistic fish lack a theory of mind? And how could you tell?
Nitpicker - Indeed. I have it on good authority that these fish were so severely autistic that they never learned to speak and, in fact, spent all their time going in circles in a tiny space :(
You let them off easy by not asking how these lists of supposed "autism genes" were derived. Most are from microarray studies of post mortem human brain, with all the attendant caveats - there are no clear findings from that literature that justify the label of a set of "autism genes". Also, their supposed finding is not corrected for multiple testing - they compare their list of genes to a long list of gene-sets (dubiously) associated with various conditions. One of these showed a statistically unlikely overlap - unlikely that it would be that one, not unlikely that it would be any of them. Finally, microarray analyses of gene expression across all the cell types in the brain are just stupid. Which cells are you expecting to see changes in? How big a change must you see in that cell to detect it when you dilute with the other hundreds of cell types? It's just bad, bad, bad! And the reporting of it is dangerous and irresponsible.
Erm... which profession? Who claimed to find it in fish? I haven't read anyone say that these fish are autistic. The scientists just do studies and produce papers. Muppets in the media interpret it all and give it a sensational title. Have you read the article itself, or are you just being all sensational and jumping to conclusions yourself? It's fair to say that these investigations do nothing other than to perhaps provoke further discussion and investigation. By the way, you spell it 'losing'. Just the one 'o'.
actually, the authors state at the end of their abstract:
"Our findings suggest a new potential trigger for idiopathic autism in genetically susceptible individuals involving an overlooked source of environmental contamination."
That is pretty rich, given how far what they did to their fish larvae is removed from anything that would ever happen to a human foetus/infant from environmental exposure. Furthermore, direct assessment of children of mothers who did take SSRIs during pregnancy, arguably leading to much higher exposure than through drinking water, have shown only a moderately increased incidence of autism (http://www.ncbi.nlm.nih.gov/pubmed/21727247).
This is sh!^. If a fish is autistic in chemical water then I guess a cockroach is autistic when sprayed with insecticides.
Why was the mother depressed in the first pace should be explored, and not what the SSRIs did.
Most medical science research is crap. So this one is fishy too and the fishy really comes only in one line at the end of the article.
Move along nothing to see here.
Neuroskeptic, you wrote:
"There were multiple fish per tank, five in fact, but those are not five independent observations, because they shared a tank."
Are you going to criticize all human studies for the lack of inter-individual independence because we all live on the same planet?
The other ridiculous thing about this study: fish are taking in water 24/7 - it is their "air!" Humans ingest drinking water at substantially lower rates. So 400-700 times the rate of drinking water + infinitely more exposure to said water in a species with a much smaller brain = their desired finding. I make the last statement because it seems to me that the authors did everything possible to show a positive "finding" and then went out of their way to make way too much of it. I can't think of more irresponsible science.
Catherina: That's actually a really good point about use of SSRIs in pregnancy, I never thought of that.
If the tiny quantities of drugs in water were causing even a little autism, you'd expect someone who actually took SSRIs to have a far higher rare of autism, because the drug levels would be 1000s of times higher.
Yet that paper found that antidepressant use during pregnancy merely doubled the risk. That's still important (if true) but as you say, it casts serious doubt on the idea that traces have an effect.
Anonymous: Are you going to criticize all human studies for the lack of inter-individual independence because we all live on the same planet?
No - because we all live on the same planet it's not a variable. It doesn't vary.
The problem with the tanks is that the fish sharing a tank are likely to be alike, just because they share a tank.
Maybe one of the tanks is slightly warmer, darker, different oxygen levels, gets fed later in the day... or maybe one of the fish in the tank is an asshole who gobbles up all the food, stressing the other fish out.
So they fish within a tank are not independent.
Hi Neuroskeptic
I have not read the study and you did not mention the distribution of fish in the anti-D tanks and the control tanks. The anti-D is the variable, not the distribution in the tanks. If the distribution of the fish in the tanks is the same for the anti-D group and the control group then proper control has been established, has it not?
Also your criticism of the environment within each tank would still hold if the fish were in separate tanks. The environment is simply something else to be controlled.
You obviously have not been pregnant Neuroskeptic. You can't take drugs during pregnancy. Even alcohol can effect your unborn. External trauma like slapping your tummy can effect your unborn. Of course SSRIs and everything else to alleviate your depression like the depressive thoughts themselves can cause a miscarriage.
I wonder what happens if you eat autistic fish.. conspiracy, like this study.
Anonymous: "Also your criticism of the environment within each tank would still hold if the fish were in separate tanks. The environment is simply something else to be controlled."
Sorry, I explained that badly. I'll try again.
Each tank will have a certain environment. If every fish had its own tank then this would, at worst, add random noise to the study.
However, since fish share a tank, it also means that the effective sample size is reduced because the fish sharing a tank are likely to be correlated with each other, and therefore are not independent data points.
For example. Suppose I wanted to do a study of the effect of putting fluoride in the water supply, on cancer.
I want my study to have 2 million people, 1 million on "drug" 1 million on placebo.
If I took 2 million people and randomly assigned them individually to get fluoride water or normal water then that would be fine.
But suppose I decided to just take 2 cities, of 1 million people each, and dump fluoride in the water supply for one of them.
That's still 2 million people but it would be wrong to treat it as a study with n=2,000,000.
Suppose I found that the people in the drug city had a 10% higher cancer risk. That could be because that city just happened to have more cancer (for whatever reason) and by chance it happened to be the drug city, and there is a 1 in 2 chance that it was.
If I did a t-test with 2 million datapoints it would come out as massively significant effect of drug but actually the chance of that happening is 50% (assuming a preexisting difference - but we can't rule that out - unless we do a pre-post study but that's a whole new kettle of, er, fish).
omg said...
I wonder what happens if you eat autistic fish..
Hilarious idea.
Now a study about more autistic mammals, we've done mice. I've seen some wild horses that i'd call pretty autistic. Didn't look me in the eye, didn't communicate much and had severe rage attacks especially when you tried to mount them. Evidently due to their autism they didn't like being touched.
My rabbits are autistic.
They never speak, they just utter the odd grunt now and then. They despise being held, and lash out.
They frequently become obsessed with certain items, usually electrical or mechanical things, and they have savant like abilities, they always know exactly what time it is, whether it's feeding time or v...e...t time. It's uncanny.
ABA is working quite well. They will engage in desirable behaviours (responding to their names, allowing my socks to survive, making a quick telephone call) when they're modelled to them.
I'm sure. this has nothing to do with the bushels of mint leaves in their therapy space. Just a happy accident that they're surrounded by their favourite food. Oh and it's GF/CF too, bonus!
Now, about my bipolar goldfishies...
Neuroskeptic wrote:
"However, since fish share a tank, it also means that the effective sample size is reduced because the fish sharing a tank are likely to be correlated with each other, and therefore are not independent data points."
What would be correlated?
Yet that paper found that antidepressant use during pregnancy merely doubled the risk.
I had a look at the paper by Croen &c -- the one cited in the fish paper -- and it turns out that they didn't find out anything about *risks*. What they calculated was the Adjusted Odds Ratio:
mothers of children subsequently diagnosed with ASD were twice as likely to have at least 1 antidepressant prescription in the year prior to delivery of the study child
The increased *risk* is unknown, except that it's much smaller than a factor of two.
The authors of the fish paper should have read Croen's paper rather than simply relying on a press release.
I am gratified that no-one else has made a joke about "social morays".
Anonymous: "What would be correlated?"
Everything, potentially. Fish sharing a tank are probably going to be more alike on any given measure you can think of, compared to fish not sharing a tank.
In this case I think they might be more alike in terms of gene expression. Being in a particular tank doesn't directly affect brain gene expression but it could do so via many mechanisms e.g. if one tank is more stressful, this will cause cortisol release, activate stress response genes in many brain areas. For example.
Smut Clyde - thank you, I should have read that better. That makes the authors' claims worse.
Still, you might want to ask the question of how much of big pharma's profits go into public water filtration? If we're pissing away birth control hormones and anti- depressants, at what point do we acknowledge that this takes place within a closed system, and that certain compounds don't break down quite as readily as we had hoped. Maybe if these longer term cleanup costs were figured in up-front, we would be less subject to the short term whimsy and greed of markets, the diagnosis-to-fit-the-drug phenomenon?
So, I have a dumb question: looking at the supplementary files, the ones that actually list each gene in each set, why is Table S3 so much larger than Tables S1 and S2 combined? Wouldn't you expect S3 to be the other two put together, but listed by signal-to-noise ratio instead of by set? Where'd all the extra genes in S3 come from??
Neuroskeptic wrote:
"Everything, potentially. Fish sharing a tank are probably going to be more alike on any given measure you can think of, compared to fish not sharing a tank."
Everything? Anything? Or nothing.
To assume that important variables are correlated without proof of such correlations is not n acceptable means of experimental design, is it? If the fish were in separate tanks as you suggested then I could come up with an equally untested mechanism by which even there might still be correlations, couldn't I?
Post a Comment